logbook.json

{
  "schema_version": 1,
  "title": "SeqDataVal — Reproduction",
  "emoji": "🎯",
  "space_id": "snaykey/repro-seq-dataval",
  "paper": {
    "openreview_id": "GFFMLwn053"
  },
  "tags": [
    "icml2026-repro",
    "paper-GFFMLwn053"
  ],
  "updated_at": "2026-07-24T23:44:59+00:00",
  "root": {
    "slug": "index",
    "title": "SeqDataVal — Reproduction",
    "file": "pages/index.md",
    "children": [
      {
        "slug": "executive-summary",
        "title": "Executive Summary",
        "file": "pages/executive-summary/page.md",
        "children": []
      },
      {
        "slug": "claim-1",
        "title": "Data selection is reformulated as a sequential decision-making problem (Definition 3.2) maximizing expected utility across all selection sizes, ∑_{k=1}^{|D|} U(S_k), with an exact Bellman-equation DP solution V(s) = U(s) + max_a V(s ∪ {a}) (Equation 2).",
        "file": "pages/claim-1/page.md",
        "children": []
      },
      {
        "slug": "claim-2",
        "title": "Theorem 4.3 shows that under linear utility functions, semi-value-based methods (Data Shapley, Beta Shapley, Data Banzhaf) achieve optimal sequential selection, while Theorem 4.6 shows that for monotonic submodular utilities with curvature c, these semi-value methods only guarantee a (1-c)^2 * OPT_k approximation, degrading quadratically as c approaches 1.",
        "file": "pages/claim-2/page.md",
        "children": []
      },
      {
        "slug": "claim-3",
        "title": "A bipartite graph-based surrogate utility is proposed that learns training-validation coverage relationships while provably preserving submodularity (Theorems G.2-G.3, Section 5, Algorithm 2).",
        "file": "pages/claim-3/page.md",
        "children": []
      },
      {
        "slug": "claim-4",
        "title": "On curvature-controlled synthetic experiments, game-theoretic valuation methods achieve mean accuracy above 0.70 at curvature 0.0 but degrade to 0.59 at maximum substitutability (curvature 1.0), empirically validating the approximation-guarantee theorem (RQ2 experiments).",
        "file": "pages/claim-4/page.md",
        "children": []
      },
      {
        "slug": "claim-5",
        "title": "The bipartite surrogate method reaches over 60% accuracy on the bbc-embeddings dataset with only 20 selected samples (versus 40-60 samples needed by baselines) and 80% accuracy on the digits dataset with just 25 samples (Figure 4, RQ3, Section 7).",
        "file": "pages/claim-5/page.md",
        "children": []
      },
      {
        "slug": "conclusion",
        "title": "Conclusion",
        "file": "pages/conclusion/page.md",
        "children": []
      }
    ]
  },
  "revision": "1784936699673917900"
}